The rising prevalence of artificial intelligence (AI), coupled with the growing demand of AI-enabled software and applications, has prompted an imperative for the development of models that are not only highly capable but also optimized and efficient in terms of their memory and computational requirements for training and usage of the model. This need is particularly pronounced in federated environments, where the objective is to reduce bandwidth usage during the transmission of models. Such optimization becomes pivotal for ensuring the viability of these models on devices characterized by limited computational capacity. Importantly, federated learning not only addresses the efficiency of model deployment but also ensures that each client receives a model tailored to the unique characteristics of its data, enables privacy of training data, and mitigates the necessity for data collection, enhancing the relevance and applicability of the learned insights. In this study, we employ stripe-wise pruning, a filter pruning technique in federated environments. The application of this technique significantly reduces bandwidth utilization and trims down model size by 83.46 %. This dual benefit of resource efficiency and privacy enhancement positions stripe-wise pruning as a compelling solution for federated environments seeking to balance performance, adaptability, and data security.